Arieke J. Rijken , Netherlands Interdisciplinary Demographic Institute
In the new round of data collection of the Gender and Generations Programme (GGP), which started in 2020, many countries decided to conduct their Generations and Gender Survey as a web-only survey. A trend driven by increased costs of face-to face interviews and lockdown measures due to the COVID Pandemic. A major challenge in conducting web surveys are break offs, which lead to incomplete surveys and potentially to breakoff bias. In this paper I use recent data from GGS web surveys in (at least) five countries to study break offs. I present descriptive results on cross-national variation in the percentage of break offs (ranging from 17 to 43%) and in break off patterns: where in the questionnaire do respondents break off? In the discussion of these results, I pay attention to country differences in implementation of the survey (e.g. how are respondents recruited, how do they reach the web survey, what are the incentives?). Furthermore, I conduct logistic regression analyses to test hypotheses on the association between respondent characteristic and breaking off. The length of the questionnaire is heavily affected by the life history and family composition of the respondent due to routing and loops. Therefore will test hypotheses on effects of, e.g. having a partner, number of ex-partners and number of children. Do all these factors which increase the length of the survey also increase the chance that a respondent breaks off? Finally, we investigate the association between device used and breaking off.
Presented in Session P1. Postercafe